SAN FRANCISCO — OpenAI opened its annual DevDay conference on Monday, introducing an expanded platform for building autonomous AI agents and announcing sharp reductions in its API pricing to developers. Chief executive Sam Altman delivered the keynote address.

The company positioned the new agent tools as a way for developers to build software that can plan and execute multi-step tasks with reduced human supervision. The updated framework includes native support for tool use, memory, and connections to external applications, extending work OpenAI previewed earlier in the year.

Pricing dominated the keynote. OpenAI announced lower per-token costs for its flagship models, a move it framed as making frontier capabilities cheaper for startups and enterprise customers. The reductions follow aggressive pricing from Google's Gemini and Anthropic's Claude lines, which have pressured OpenAI's market position among developers.

Analysts at Gartner have noted that competition among foundation-model providers has shifted from raw benchmarks toward cost efficiency and agentic reliability. OpenAI's announcements reflected that shift, emphasizing lower latency, cheaper inference, and tools aimed at production deployment rather than experimentation.

Altman told the audience that the company's priority was helping developers ship agents that work reliably at scale, and said further model and pricing updates would follow before the end of the year.